1. Field of the Invention
The present invention relates to a method for renewing dictionary image data in a fingerprint identification device, and particularly to a method for renewing dictionary images in a fingerprint identification device, in which dictionary identification time can be shortened.
The present invention relates to a method for renewing dictionary images in a fingerprint identification device, and particularly to a method for renewing dictionary images in a fingerprint identification device, in which the dictionary images can be kept in the best condition.
The present invention relates to a dictionary registration and identification method for a fingerprint identification device, and particularly to a multiple-dictionary registration and identification method for a fingerprint identification device, in which an identification rate is improved by registering and identifying a plurality of dictionaries.
The present invention relates to a method for judging fingerprinting in a fingerprint identification device, and particularly to a fingerprinting judging method in a fingerprint identification device, in which continuous identification rejections can be prevented.
The present invention relates to a fingerprint identification device which can prevent misjudgment in identifying a fingerprint and to a fingerprinting judging method.
2. Description of the Related Art
In a fingerprint identification device, by identifying an entered fingerprint image with fingerprint dictionary data previously registered in a dictionary memory to judge whether or not the entered fingerprint image corresponds to a finger of a particular person whose dictionary data is previously registered in the fingerprint identification device.
In such a case, it is desired that time required for dictionary identification can be shortened by enabling to quickly identify the conformity with the multiple dictionary data to be used for identification.
The fingerprint dictionary data is image data of multiple characteristic parts, such as ending points and bifurcation points, of a fingerprint image extracted by a sensor and registered together with its coordinate data, and when the entered fingerprint image is detected to conform with some of such characteristic data, it is determined to be identification pass.
On the other hand, in fingerprinting by using a sensor, it has been experienced that fingerprints are relatively easily deformed depending on the position of fingers on a fingerprinting surface, the magnitude of pressing a finger against the fingerprinting surface or the like, and the fingerprint image to be taken into the fingerprint identification device is varied.
Therefore, if dictionary data is registered as many as possible on the same person, various cases can be covered even when the fingerprints are deformed as described above. Adopting a method of determining as an identification pass when the entered fingerprint image matches with any one of dictionary images, an identification match rate (acceptance rate) of the fingerprint identification device can be improved.
FIG. 14 shows the configuration of conventional dictionary data, in which (a) indicates all dictionary data, (b) dictionary data on one person within all dictionary data, and (c) one characteristic data in the dictionary data on one person.
All dictionary data comprises dictionary data on M to-be-identified persons (1 to M), for example. Dictionary data on one person comprises N minutiae (minutiae 1 to N), for example. And, one minutia data comprises image data on minutiae and its coordinate data in a prescribed window.
In the fingerprint identification device, by using dictionary data having N minutia data registered therein, pattern matching is made from minutia 1 in order with respect to the entered fingerprint image which is taken in at the fingerprint identification, and if a pattern matches on, for example, L minutia data, it is determined as identification pass.
FIG. 15 shows an example of identifying minutiae in which (a) indicates Example 1, and (b) Example 2. Among N minutiae, those with matched patterns are indicated by .smallcircle., and those without matched patterns by .DELTA..
Example 1 shows that minutia 2 does not match but other minutiae match, and it is determined to be identification pass based on the pattern matching of (L+1) minutiae. And, Example 2 shows that minutiae 2, 3 and 5 do not match but other minutiae match, and it is determined to be identification pass based on the pattern matching of (L+3) minutiae.
Besides, in the fingerprint identification device, to complete the identification with the dictionary data soon, when N minutiae have K minutiae which do not have matched patterns in succession, it is judged that the entered fingerprint image and the dictionary data are not of the same person, it is determined as identification rejection, and identification of the remaining minutiae with the dictionary data is canceled.
FIG. 16 shows the cancellation of the identification process, in which dictionary data comprising N minutiae is identified in succession, K (=3) continuous identification mismatches are detected at minutiae 5 through 7, and the process is terminated at this point as identification rejection. In this case, even if it is possible that there may be a minutia having a matched pattern among, for example, minutia 8 through minutia N, the remaining process is canceled.
In identifying the minutiae shown in FIG. 15, when minutia data that have patterns which often do not match with entered fingerprint images are contained many in minutia data of less than 1, time required before patterns are matched with L minutia data becomes longer than when such minutia data are not contained.
And, in the cancellation of the identification process shown in FIG. 16, when the minutia data that have patterns which do not match with the entered fingerprint images are contained at least K in succession in the neighborhood of the head of the N minutia data string, an identification match rate (acceptance rate) is lowered.
FIG. 1 shows the configuration of a conventional fingerprint identification device to which the present invention are applied and which comprises a core section 10 and a control 20. The core section 10 performs registration and identification of fingerprint images, and the control 20 controls the processing in the core section 10. The control 20 sends a command to the core section 10 and receives from the core section a response to the command.
In the core section 10, reference numeral 11 denotes a central processing unit (CPU) for controlling the processing of registration and identification of fingerprint images. And, reference numeral 12 denotes a sensor for taking the image of a fingerprint which is to be registered or identified, 13 a frame memory for storing the image taken from the sensor 12 as a multivalued image, 14 an image processing circuit for processing various images, such as binarization of a multivalued image, extraction of minutiae from a binarized image, pattern matching of a binarized image with minutiae and the like, 15 a binary memory for storing a binarized image, 16 a dictionary memory for storing the minutiae extracted from a binarized image as a dictionary image, and 17 an identification memory for keeping the minutiae of fingerprints to be used for identification; and these are connected to a bus 18 of the CPU 11. And, reference numeral 19 is a control interface for interfacing of a command and a response between the bus 18 and the control 20.
As a dictionary registration identifying method in such a fingerprint identification device, use of a plurality of dictionary images has been proposed to improve an identification rate.
The dictionary images are minutiae such as ending points and bifurcation points extracted from the image of a fingerprint pressed on the sensor, and when some of such characteristics parts are detected to match with the entered image, it is determined to be identification pass.
On the other hand, in fingerprinting by using a sensor, it has been experienced that fingerprints are relatively easily deformed depending on the position of fingers on a fingerprinting surface, the magnitude of pressing a finger against the fingerprinting surface or the like, and the fingerprint image to be taken into the fingerprint identification device is varied.
Therefore, if dictionaries are registered as many as possible on the same person, various cases can be covered even when the fingerprints are deformed as described above. Adopting a method of determining as an identification pass when the entered fingerprint image matches any one in the dictionaries, an identification rate of the fingerprint identification device can be improved.
FIG. 17 shows a flowchart of registration and identification processing in a conventional multiple dictionary method, in which (a) indicates the registration processing, and (b) the identification processing.
In the multiple dictionary method, the registration processing is performed as follows.
1 When a fingerprint is taken, a fingerprint image (first image) entered through the sensor 12 is binarized in the image processing circuit 14 through the frame memory 13, and stored in the binary memory 15. PA0 2 From the binarized image stored in the binary memory 15, minutiae are extracted in the image processing circuit 14 by running the minutia extracting program in the CPU 11 to temporarily register as a fingerprint image in the identification memory 17. PA0 3 When a fingerprint is taken again, a fingerprint image (second image) is binarized in the image processing circuit 14 through the sensor 12 and the frame memory 13, and stored in the binary memory 15. PA0 4 By running the identification processing program in the CPU 11, pattern matching is performed in the image processing circuit 14 in view of the dictionary image (first image) temporarily registered in the identification memory 17 and the binarized image (second image) stored in the binary memory 15. PA0 5 When it is determined to be identification pass as a result of identification, the dictionary image for the first image temporarily registered is formally registered as dictionary 1 in the dictionary memory 16. PA0 6 Besides, from the binarized image of the second image stored in the binary memory 15, the dictionary image, which is obtained by extracting the minutiae in the image processing circuit 14 by running the minutia extracting program in the CPU 11, is formally registered as dictionary 2 in the dictionary memory 16. PA0 1 By fingerprinting, a fingerprint image (identification image) entered through the sensor 12 is binarized in the image processing circuit 14 through the frame memory 13 and stored in the binary memory 15. PA0 2 According to a command from the control 20, the dictionary image of dictionary 1 as the dictionary number which is designated from the dictionary memory 16 is transferred to the identification memory 17, and the matching program in the CPU 11 is performed to make pattern matching in the image processing circuit 14 by the binarized image in the binary memory 15 and the dictionary image in the identification memory 17. PA0 3 When the identification is resulted to be identification rejection, the dictionary image of dictionary 2 is used to perform again pattern matching with the binarized image of the binary memory 15 in the image processing circuit 14. PA0 4 When the processing in 2 above results in identification acceptance with respect to the dictionary image of dictionary 1 or the processing in 3 above results in identification acceptance with respect to the dictionary image of dictionary 2, the result is noticed as identification success to the control 20. PA0 5 When the processing in 2 and the processing in 3 are resulted in identification rejection, the results are noticed as identification failure to the control 20. PA0 (1) Changes over time (age, external injury on fingers) PA0 (2) Changes of placing fingers due to practice PA0 (1) The presence of fingerprinting is judged in the first step of the identification processing. PA0 (2) When the fingerprinting has been made in the first step of the identification processing, it is demanded to separate the finger. PA0 (3) After issuing the demand for the finger separation, the presence of fingerprinting is judged. PA0 (4) When it is checked in both (1) and (2) above that the fingerprinting has not been made, a demand for fingerprinting is issued. PA0 (5) After demanding the fingerprinting, the presence of fingerprinting is judged again. PA0 (6) When it is confirmed that fingerprinting has been made as a result of judging the fingerprinting again after demanding the fingerprinting in (5) above, the fingerprint image is captured.
Thus, when the first image and the second image have been registered formally, it is determined that the registration is success, and the processing is terminated. On the other hand, when it is determined as identification rejection as a result of identification, the processing so far is cleared as a registration failure.
By the multiple dictionary method, the identification processing is performed as follows.
By adopting the multiple dictionary method into the fingerprint identification device, identification rejection based on the deformation of fingerprints caused at fingerprinting can be prevented, and an identification rate can be improved.
However, the fingerprint is not always invariable but varies due to the following various causes.
Therefore, the fingerprint image entered every time varies gradually with respect to the previously registered dictionary image, and when fingerprinting is made later, it may result in identification rejection. Therefore, it is hard to keep a high identification rate unless the registered dictionary image is renewed at an appropriate opportunity. But, a method of renewing the dictionary image which enables the renewal of the dictionary image has not been proposed.
Another dictionary registration and identification method is known described in, for example, Japanese Patent Laid-Open Publication No. Hei 7-271980.
FIG. 18 shows the outline of a fingerprint identification device to which prior art and the present invention are applied, in which reference numeral 61 denotes a transparent prism type finger stand, 61a a top thereof, 61b an inclined face, and 61c a bottom face thereof. Reference numeral 62 denotes a prism face on which a finger is positioned at the top face of the finger stand 61, and 63 a finger positioned on the prism face. Reference numeral 64 denotes a lamp to illuminate the finger 63 from below the bottom face 61c of the finger stand 61. Reference numeral 65 denotes an image pickup device to take a picture of the finger 63 from the side of the inclined face 61b of the finger stand 61. Reference numeral 66 denotes a fingerprint identification control for controlling the lamp 64 and the image pickup device 65 to identify the fingerprint. In the fingerprint identification control 66, reference numeral 67 denotes an image processing unit for processing the fingerprint image taken from the image pickup device 65, 68 a fingerprint registering unit for registering the collected fingerprint images as the dictionary, and 69 an identifying unit for identifying the pictured fingerprint image with the registered dictionary images to judge matching.
The prism face 62 of the finger stand 61 is illuminated through the finger stand 61 from below the bottom face 61c by the lamp 64. The finger 63 is put on and pressed against the prism face 62 to contact fingerprint ridges to the prism face 62. When this sate is viewed from the inclined face 61b of the finger stand 61, the fingerprint ridges have a light reflection factor different from other parts, so that the fingerprint image formed on the prism face 62 can be pictured from the inclined face 61b of the finger stand 61 by the image pickup device 65 such as a TV camera.
The pictured fingerprint image is processed by the image processing unit 67 in the fingerprint identification control 66 to make it easy to compare minutiae, and identified by the identifying unit 69 with the minutiae of the fingerprints registered in the form of the dictionary image in the fingerprint registering unit 68. And, when a prescribed number or more of minutiae are detected to match with the entered fingerprint image, it is judged as identification pass. The minutiae of the fingerprint are extracted in view of the ending points and bifurcation points of the fingerprint ridges.
Before starting the identification of a fingerprint, it is necessary to make sure that the finger 63 is on the finger stand 61. Therefore, a fingerprinting judging process is performed to judge that fingerprinting has been made.
FIG. 19 shows the flowchart of a conventional fingerprinting judging process in the fingerprint identification device. In identifying the fingerprint, the device first requires that the fingerprint is affixed. After determining that the first fingerprinting has been made, a fingerprint image is taken and identified with the registered fingerprint images. When the identification is resulted in identification pass, processing as identification success is performed, and, for example, prescribed processing such as permission for entrance is made.
When the first fingerprinting results in identification rejection, fingerprinting is required again, and a fingerprint image is taken to perform identification processing. When it results in identification rejection, the same procedure is repeated, and if the third identification processing results in identification rejection, processing as identification failure is performed, and, for example, processing such as prohibition against entrance is made.
In the conventional fingerprinting judging process shown in FIG. 19, when the identification results in identification rejection, fingerprinting is required again and a fingerprint image is taken accordingly, the finger on the finger stand remains as it is and kept in close contact with the fingerprinting surface (prism face 2) through the three identifications.
When the first identification results in identification rejection and fingerprinting is required again by the fingerprint identification device, if the finger is continuously in close contact with the fingerprinting surface from the first fingerprinting, the fingerprint image taken upon another requirement for fingerprinting is almost the same as the fingerprint image taken first. Therefore, when the first-taken fingerprint image results in identification rejection, it is highly possible that the second-taken fingerprint image also results in identification rejection.
Therefore, in the second fingerprinting, if it is urged that the finger is separated from the fingerprinting surface after the first fingerprinting and then positioned on the fingerprinting surface for fingerprinting again, it may be highly possible that the fingerprint image taken by the second fingerprinting results in identification pass, but such a method has not been adopted.
And, when the finger is positioned on the fingerprinting surface before the first fingerprint image collection and kept thereon through a plurality of fingerprint image collections in the same way as by the conventional method, fingerprinting time becomes long, the finger skin stretches or shrinks to deform the fingerprint image, and an identification rate is lowered. Besides, when the finger on the fingerprinting stand is not in position, the fingerprint is partly deformed, fingerprinting time becomes long, the fingerprint image is heavily deformed, and an identification rate is lowered.
Since the fingerprint image is taken in immediately after checking the fingerprinting by the device, it is felt that the fingerprint image is taken immediately when the finger is put on the fingerprinting surface. Therefore, when the finger is separated from the fingerprinting surface after each fingerprinting as in the method of the present invention, timing for taking the fingerprint image after fingerprinting can be kept constant and it can be taken in a short time after fingerprinting, so that the identification rate at the first fingerprinting and that at the second fingerprinting can be kept same by providing the same conditions for taking fingerprints. But, such a method has not been proposed.
FIG. 20 is an explanatory view of the fingerprint image pickup unit, in which reference numeral 71 denotes a camera, 72 a light source, 73 a prism, 74 an elastic film, and 75 a finger subject to fingerprinting. The camera 71 can be a CCD camera, for example. When it is provided at an angle of critical angle .theta. or below with respect to the fingerprinting surface (the surface of the elastic film 74) and the fingerprinting surface is illuminated through the prism 73 by the light source 72, the reflected light does not enter the camera 71 if the finger 75 is not in position. And, when the finger 75 is put on the elastic film 74 on the fingerprinting surface, the reflected light which corresponds to the ridge and valley lines forming the fingerprint of the finger 75 enters the camera 71, and a signal of the pictured fingerprint is outputted from the camera 71.
And, the elastic film 74 serves to improve the adhesion with the finger 75 and to enable to take a clear picture of the fingerprint of even a dry finger, and is required to have abrasion resistance to some extent. And, there is not any colorless transparent material available. Accordingly, when external intense light enters the fingerprinting surface, scattered light may be produced by the elastic film 74 and entered into the camera 71. And, if the elastic film 74 is scratched or made dirty, reflected light is produced by the light from the light source 72 and entered into the camera 71.
In the fingerprint identification device, the finger 75 is positioned on the fingerprinting surface (the elastic film 74) and the fingerprint is pictured by the camera 71. And it is necessary to judge whether or not the image signal from the camera 71 is derived from the fingerprint image. Therefore, in the conventional fingerprint identification device, for example, a method (a) is known in which when the finger is positioned on the fingerprinting surface, the reflected light from the finger enters the camera 71, average luminance within a judging area used to identify the fingerprint in the pictured screen by the camera 71 is determined, and when the average luminance exceeds a predetermined threshold, it is presumed that the reflected light from the finger 75 has entered the camera 71, it is judged that fingerprinting has been made, and the fingerprint image signal at the time is taken to make the registration or identification of the fingerprint.
A method (b) is also known in which the judging area is divided into a plurality of blocks, and when the number of blocks which have the average luminance of respective blocks exceeding a threshold exceeds a prescribed value, it is judged as a fingerprinting state. And, as to the number of blocks in the judging area, when it is assumed that the number of blocks in the judging area is j, the number of picture elements in the block is m, and luminance of the picture elements in the judging area is Pqn, average luminance A q of a certain block is expressed as follows. ##EQU1## And, a method is also known in which a total of blocks satisfying the average luminance A q larger than a certain threshold a is determined and regarded as N, and when the total N becomes larger than a certain threshold b, it is judged that fingerprinting has been made.
In the conventional method (a) that it is judged that fingerprinting has been made when the average luminance in the judging area exceeds the threshold, if intense light such as direct sunlight enters the fingerprinting surface, the scattered light from the elastic film 74 enters the camera 71 to make the entire picture white, and the average luminance is increased. In other words, an error is caused in judging the fingerprinting.
And, in the conventional method (b) that it is judged that fingerprinting has been made when numeral N indicating the number of times that the average luminance of respective blocks in the judging area exceeds the threshold a is determined and exceeds the threshold b, lighting nonuniformity can be compensated, but in the same way as the conventional method (a), the average luminance of respective blocks is increased when intense light enters the fingerprinting surface, an error occurs in judging fingerprinting.